Paper R. Meyer Brand 2013
Ralf M. Meyer, Thomas Brand, Acta acustica united with Acustica 99(3), 442-456, May/June 2013
This study compares three different extensions of the Speech Intelligibility Index (SII, ANSI S3.5-1997) designed for calculating speech reception thresholds (SRTs) in fluctuating noise. The extensions differ in the amount of temporal information taken into account. They are used to predict individual SRTs of listeners with normal and impaired hearing in two experiments. The three extensions take into account different assumptions about the fluctuations: (1) broadband overall level fluctuations of the noise, (2) level fluctuations of the noise analysed in different frequency bands, and (3) level fluctuations of speech and noise analysed in different frequency bands. Experiment I (12 listeners) used stationary speech shaped noise, sinusoidally modulated noise, rectangular modulated noise and focused on the test of the implementations. Experiment I showed the limits of the model versions and verified our implementations. Experiment II (113 listeners) used noise with speech-like modulations. The third model extension could explain the highest proportion (52%) of the variance of the large database of experiment II. However, the additional complexity of the second and third extension yielded only small improvements compared to the first extension.